The comprehension of safety signs for construction workers was compared among existing and improved designs. Safety signs considered were existing signs, existing signs improved by adding text, existing signs improved by adding/changing pictorials, and existing signs improved by adding/changing pictorials and text. Three hundred and eighty-four construction workers were randomized for assignments into four groups. They were working on nine constructions work sites randomly selected from work sites located in Bangkok, Thailand. The number of workers in each work site was calculated for the proportion of the sample size and then they were randomized so that equal numbers of workers were included in the four groups. The four types of safety signs and an interview form were used to test their comprehension of safety signs. For data analysis, percentages, means, standard deviations, the chi-square test, the one-way ANOVA and LSD tests were applied. The results showed that the mean comprehension scores of the existing signs improved by adding text, the existing signs improved by adding/changing pictorials, and the existing signs improved by adding/changing pictorials and text were significantly higher than those of the existing signs. In addition, the mean comprehension scores of the existing signs improved by adding text and the existing signs improved by adding/changing pictorials and text were significantly higher than those of the existing signs improved by adding/changing pictorials. The highest mean comprehension score of safety signs for the construction workers was found in the existing signs improved by adding/changing pictorials and text. These results indicated that for construction workers, the comprehension of safety signs with pictorials and explanatory text was better than that of the existing signs.